Abstracts

Session I: Observed Oceanic Decadal Variability

Decadal Climate Prediction: Challenges And Opportunities

J. Hurrell (NCAR)

The scientific understanding of climate change is now sufficiently clear to show that climate change from global warming is already upon us, and the rate of change as projected exceeds anything seen in nature in the past 10,000 years. Uncertainties remain, however, especially regarding how climate will change at regional and local scales where the signal of natural variability is large. Decision makers in diverse arenas, from water mangers in the U.S. Southwest to public health experts in Asia, need to know if the climate events they are seeing are the product of natural variability, and hence can be expected to reverse at some point, or are the result of potentially irreversible anthropogenic climate change. The climate science community will not be able to answer these questions and reduce the uncertainties in near-term climate projections without moving toward high resolution climate system predictions, with a blurring of the distinction between shorter-term predictions and longer-term climate projections. The key is the realization that climate system predictions of natural and forced change, regardless of timescale, will require initialization of coupled general circulation models with the best estimates of the current observed state of the atmosphere, oceans, cryosphere, and land surface, a state influenced both by the current phases of modes of natural variability and by the accumulated impacts to date of anthropogenic radiative forcing. Formidable challenges exist: for instance, what is the best method of initialization given imperfect observations and systematic errors in models, what effect does initialization have on climate predictions, what predictions should be attempted and how would they be verified? Accurate initial conditions for the global oceans are especially important and could conceivably be provided by ARGO floats and existing ocean data assimilation exercises. However, performing hindcasts prior to the ARGO float era of near-global upper ocean salinity and temperature data will remain a significant problem and possibly compromise the development of prediction capacities. Despite such challenges, it needs to be recognized that useful predictions of the evolution of the climate system over the next few years to decades requires initialized projections that exploit the predictive potential contained in both past and future changes in radiative forcing and in slowly evolving phenomena, such as ocean current systems and heat content, that give rise to internal decadal variability. Climate predictions that exploit the full predictive potential of forced and free climate variations will form the raw material for the decision making needed to enable societies to adapt to the climate changes of the next few years to decades.

What Do Observations Tell Us About The Changing MOC?

S. Cunningham (NOC)

Observations of relevance to measuring MOC change fall into four classes: 1. Estimates of the MOC based on measurement and inference of the full-depth, basin wide circulation; 2. Processes implicated with or indicative of, MOC change; 3. Recent climate signals possibly associated with MOC change and; 4. Evidence for substantial past changes in the MOC that extend to a broad influence on climate.

In this talk I focus on the first two classes. I will briefly review traditional hydrographic estimates of the MOC, then focus on dynamics and heat flux estimates from the 26.5°N RAPID/MOCHA basin wide mooring array.

Overflows of dense water across the Greenland-Iceland-Scotland Ridge and formation of Labrador Sea Water are implicated in MOC change: I will discuss the latest observations of these processes, particularly those being made through the European THOR project.

Recently, direct association of the MOC with thermohaline forcing has motivated reconstructions of the mid-latitude MOC at 48°N based on surface heat and freshwater fluxes alone. The surface-forced MOC inferred using this approach has reduced from the 1990’s to the early 2000’s by approximately 3 Sv. At 42°N altimeter and subsurface float data have been used to estimate the time-averaged circulation in the upper ocean from 2002 to 2009, and show no trend over this period.

Arctic Sea Ice Predictability

M. Holland (NCAR)

Observed Decadal Variability In The Indo-Pacific Sector

C. Deser (NCAR)

In this overview, I will discuss the observational record of low-frequency variability in the Indo-Pacific region based on a wide variety of indicators including SST, air temperature, cloudiness, precipitation, and sea level pressure. I will compare the spatial distributions of tropical and extra-tropical climate anomalies associated with interannual (ENSO) and low-frequency (bi-decadal and ~50 year) variability. Finally, I will present some of the leading paradigms for the decadal-scale fluctuations, including the roles of oceanic mixed layer processes and dynamics.

Decadal climate variability in and around the Pacific: causes and predictability - S. Power (BoM), S. McGregor (IPRC), G. Kociuba (BoM), J. Callaghan (BoM), N. Holbrook (Uni. Tasmania) and D. Rodriguez (QDPIF)

Decadal Climate Variability (DCV) is very important to life in and around the Pacific. It is present in e.g. rainfall, river flow, crop yield, sea-level and the frequency of land-falling tropical cyclones. A quick survey of this variability will be provided, and some of the impacts this variability has on decision-making in agriculture will be outlined. The DCV includes remarkably large decadal variability in ENSO and ENSO teleconnections, some of which is linked to decadal ENSO-like patterns of variability like the Interdecadal Pacific Oscillation and the Pacific Decadal Oscillation. The origins of these patterns and the robust statistical links between ENSO-like decadal patterns and ENSO teleconnections will be discussed and the predictability of generational changes in ENSO teleconnections will be assessed. The physics underpinning multi-year predictability in the tropical Pacific, and reasons why ENSO indices have exhibited interdecadal and longer-term changes over the past century will be examined.

References

Callaghan, J. and S. Power, 2010: Variability and decline in the number of severe land-falling tropical cyclones over eastern Australia since the late 19th century. Climate Dynamics, (in press).

McGregor, S., N.J. Holbrook, and S.B. Power, 2009: The response of a stochastically forced ENSO model to observed off-equatorial wind-stress forcing. J. Climate, 22, 2512-2525.

Power, S.B., and G. Kociuba, 2010: Impact of global warming on the SOI. Climate Dynamics (submitted).

Power, S.B., and I.N. Smith, 2007: Weakening of the Walker Circulation and apparent dominance of El Nino both reach record levels, but has ENSO really changed? Geophys. Res. Lett., 34, doi: 10.129/2007/GL30854.

Power, S.B., and R. Colman, 2006: Multi-year predictability in a coupled general circulation model. Climate Dynamics, 26, 247-272.

Power, S., M. Haylock, R. Colman, and X. Wang, 2006: The predictability of interdecadal changes in ENSO and ENSO teleconnections. J. Climate, 8, 2161-2180.

Rashid, H., S.B. Power, and J. Knight, 2010: Impact of multi-decadal fluctuations in the Atlantic thermohaline circulation on Indo-Pacific climate variability in a coupled GCM. J. Climate (in press).

The Many Flavours Of Southern Ocean Variability Determined From Limited Observations

B. Sloyan (CSIRO)

The Challenges Of Using Ocean Modeling To Predict Coral Reef Ecosystem Response

J. Kleypas (NCAR)

Coral reef ecosystems occupy a narrow band of the sea floor in waters warmer than about 16C, and shallow enough (< 100 m) to allow photosynthesis adequate for coral and algal growth.  Thus, despite their obvious importance to marine biodiversity, fisheries, and coastal protection, the areal extent of coral reefs is a mere fraction of the open ocean, and almost all occur in coastal regions where physical oceanographic processes are not adequately captured by most numerical models.  This talk outlines the limitations (i.e., "challenges") of predicting the responses of coral reef ecosystems to climate change and ocean acidification.

What Is Causing The Decadal Warming Trend In Antarctic Bottom Water In The Abyssal Atlantic Ocean?

M. P. Meredith (BAS), A. C. Naveira Garabato (NOC), A. L. Gordon (Lamont Doherty Earth Observatory), P. Abrahamsen (BAS), L. Jullion (NOC), G. C. Johnson (PMEL)

Antarctic Bottom Water (AABW) in the abyssal Atlantic Ocean has been undergoing distinct warming in recent decades, with significant implications for calculations of the Earth's energy budget and sea level rise. Understanding the cause of this warming is a key first step towards developing predictive skill concerning how this water mass may evolve in the future, and in determining processes necessary for inclusion in coupled climate models that may ultimately seek to reproduce such changes. Analysis of temporally-sparse hydrographic sections in the eastern Scotia Sea led to an initial hypothesis, namely that the wind-forced baroclinicity of the Weddell Gyre controls the northward penetration of AABW from the Weddell Sea across the rough topography of the South Scotia Ridge and into the lower limb of the Atlantic overturning. Recent analyses of AABW properties on a repeat hydrographic section in Drake Passage (the most frequently-occupied section in the Southern Ocean) lend support to this hypothesis, and also provide information on the nature of the wind-forcing of the Weddell Gyre. Coupled modes of climate variability, in particular the Southern Annular Mode (SAM), appear significant in controlling the cyclonic wind forcing over the Gyre, and hence the baroclinicity of the gyre circulation and AABW export. Temperature records from moorings within the boundary current of the Weddell Sea and immediately downstream of the AABW outflow in the Scotia Sea provide further support for this theory, and also provide (for the first time) information on the degree of temporal variability of AABW outflow speed in response to changing cyclonic winds. We discuss the relevance of these processes to the observed AABW warming trend at lower latitudes, possible future changes in AABW given our current knowledge of the evolution of coupled climate modes at high latitudes, and ongoing efforts to monitor and better understand the AABW changes and the key processes involved.

Forced Versus Intrinsic Variability Of The Kuroshio Extension System On The Decadal Timescales

B. Qiu (University of Hawaii at Manoa), S. Chen (University of Hawaii at Manoa), and N. Schneider (University of Hawaii at Manoa)

A ubiquitous feature of the northern hemisphere subtropical ocean circulation is the the existence of an anticyclonic recirculation gyre (RG) on the southern flank of the wind-driven western boundary current outflow. The RGs significantly enhance the eastward volume and heat transport of the western boundary currents and their variability has been recognized in recent years to be crucial in understanding the decadal midlatitude oceanic changes. The dynamic cause for the decadal WBC variability, especially whether the observed variability is externally forced or reflects the intrinsic, nonlinear RG behavior, is still under debate. This cause in the Kuroshio Extension (KE) system is examined in this study by analyzing satellite altimeter sea surface height data, adopting simplified dynamic models, and evaluating nonlinear eddy-mean flow interaction. Long-term SSH measurements reveal clearly that the KE system oscillates between a stable and an unstable dynamic state. Transitions between the two states are caused by PDO-related, basin-scale wind stress curl forcing in the eastern North Pacific. During the positive PDO phase, wind-induced negative SSH anomalies from the east work to weaken the KE jet, shifting its path southward. The latter migration causes the KE jet to override the shallow Shatsky Rise, leading to an enhanced eddy kinetic energy (i.e., unstable) state of the KE. With a time lag of 1~2 yrs, the enhanced eddy variability is found to strengthen the KE's southern RG. Helped by the incoming, positive SSH anomalies from the east due to the negatively-phased PDO forcing, this eddy-driven circulation works to switch the KE system to a dynamically stable state. While the eddy-driven nonlinear RG dynamics facilitates modulations of the KE system, the decadal transitions of the KE system are found to be largely paced by the external PDO wind forcing. Specifically, we argue that the decadal timescale in the PDO wind forcing originates in the coupling of the KE SST variability and the basin-wide, midlatitude atmospheric responses.

Session II: Decadal Climate Variability and the Role of the Ocean

The Atlantic Meridional Overturning Circulation And Climate - Variability, Predictability And Change
T. Delworth, R. Zhang, S. Zhang, T. Rosati, K. Dixon, R. Msadek, F. Zeng, H. C. Lee, W Anderson (NOAA GFDL)

Research over the last decade has shown the large-scale climatic impact of Atlantic SST fluctuations on climate. These include Sahelian and Indian monsoonal rainfall, Atlantic tropical storm activity, summer climate over North America and Europe, as well as impacts on the Pacific and hemispheric temperature. Here we briefly synthesize those impacts and their relationship to the Atlantic. A crucial question is whether such variability is unique to the period of the instrumental record. To address this we review evidence from paleo records which suggest that such Atlantic variability has been present for at least the past several centuries, and perhaps much longer. Models have provided an important perspective on this variability, its origins and climatic impacts. We examine a 4000 year control simulation of the GFDL CM2.1 model, and show that the simulated AMOC has variability on both the interdecadal and multicentennial time scale. While the interdecadal variability is largely confined to the North Atlantic, the multicentennial variability encompasses the entire Atlantic, and appears to be related to out of phase variations in deep mixing between the North Atlantic and the circum-Antarctic. This variability is a major driver of  simulated variability in Northern Hemisphere temperature on multicentennial time scales. Finally, since our assessments of the impact of ocean variability on climate are largely derived from models, we present preliminary results from a newly developed global climate model (GFDL CM2.5) with significantly enhanced spatial resolution (10-25 Km in the ocean, 50 Km in the atmosphere).

Decadal Prediction For The Atlantic

J. Marotzke (MPI-M)

North Pacific And North Atlantic Multidecadal Variability: Origin, Predictability, And Implications For Model Development

M. Latif (IFM GEOMAR)

The North Pacific and North Atlantic sea surface temperature (SST) record of the last about 150 years displays strong multidecadal variability (Fig. 1). Similar multidecadal variability is also seen, for instance, in Northwest American surface air temperature (SAT), Sahel rainfall or Atlantic hurricane activity. The existence of the multidecadal variability makes climate change detection a challenge, since Global Warming evolves on a similar timescale. The ongoing discussion about a potential anthropogenic signal in the Atlantic hurricane activity is an example.

Fig. 1: Top panel: Pacific Decadal Oscillation, as denoted by annual SST based on the leading EOF SST pattern for the Pacific basin north of 20°N, updated from Mantua et al. 1997. The color fill in both the upper and lower panels is from a low-pass symmetric filter with 13 total weights and a half-power point at 16 year periods, with the end points reflected. Bottom Panel: Atlantic Multidecadal Oscillation, as denoted by the low-pass filtered time series of annual SST anomalies averaged over the North Atlantic (0-60°N, 0-80°W) relative to 1901 to 1970 (°C). The color fill depicts the low-pass filtered SST, and the solid lines are annual means. Updated from Trenberth and Shea (2006). From Hurrell et al. 2010.

A lot of work was devoted during the last years to understand the dynamics of the multidecadal variability, and external (solar forcing and aerosols) as well as internal mechanisms were proposed. The talk focuses on three aspects. First, it describes potential internal (stochastic) mechanisms for the generation of North Pacific and North Atlantic multidecadal variability. Stochastic variations in the major atmospheric pressure systems (the Aleutian low and the North Atlantic Oscillation (NAO)) and the associated surface fluxes of heat, freshwater and momentum drive low-frequency changes of the large-scale ocean circulation in the two basins. While in the North Pacific, it is the (wind-driven) circulation specifically the gyre adjustment that plays a key role, it is the (thermohaline) meridional overturning circulation (MOC) in the North Atlantic which integrates the stochastic atmospheric forcing. Different forms of stochastic models, however, were proposed, ranging from a simple AR1-process to stochastically driven coupled air-sea modes (Fig. 2).

In the North Pacific, the multidecadal variability seems to be forced by processes internal to the region. This is demonstrated by a multi-millennial control integration of the Kiel Climate Model (KCM) which simulates rich North Pacific variability on a wide range of timescales, from interannual to multidecadal and even longer. While for variations up to decadal remote forcing from the Tropical Pacific appears to play a strong role, this is no longer the case for the multidecadal variability.

Fig. 2: Different types of stochastic climate models. From Latif et al. 2002.

In the North Atlantic, a picture consistent with a stochastically driven oceanic eigenmode may serve as null hypothesis for the generation of multidecadal variability (Fig. 2, middle panel). The variability of the Atlantic Meridional Overturning Circulation (AMOC) is a key player in the generation of the multidecadal variability. Results from an NCEP-forced ocean general circulation model (NEMO, 0.5°) simulation show a consistent evolution of the AMOC in association with decadal-scale changes of the North Atlantic Oscillation (NAO), with variations of the AMOC lagging those of the NAO by approximately a decade (Fig. 3).

Fig. 3: The overturning stream function simulated in a forced integration with an ocean general circulation model (NEMO with a horizontal resolution 0.5°). Shown are the reconstruction from the leading (multidecadal) SSA mode for the years 1975, 1980, 1995, and 2000. From Alvarez et al. 2008.

The ocean model results are supported by data (Fig. 4). Low frequency-variations of the NAO drive anomalous heat fluxes over the Labrador Sea, which in turn force variations of Labrador Sea convection and subsequently variations of the AMOC. The latter were “reconstructed” by computing a dipolar SST index (North Atlantic minus South Atlantic), as AMOC variations strongly project on this index in climate model simulations. However, the role of the Tropics in the generation of multidecadal variability in the North Atlantic is still unclear and needs further consideration.

Fig. 4: The North Atlantic Oscillation (NAO) index, a dipolar SST index defined as the temperature difference between the North and South Atlantic, and a measure of Labrador Sea convection (all three low-pass filtered). The hypothesis is that low-frequency changes of the NAO drive through changes in Labrador Sea convection low-frequency changes in AMOC strength. The latter is “defined” in terms of the Atlantic dipole SST index. From Latif et al. 2010.

Second, the climate predictability in the North Pacific and North Atlantic Sectors is discussed. Several studies indicate a relatively high decadal predictability potential in these regions. Interestingly, and in contrast to seasonal to interannual predictability, potential decadal predictability is found predominantly over the mid to high-latitude oceans. The two indices displayed in Fig.1, for instance, contain rather strong contributions from the mid latitudes.

Finally, third, the paper discusses the biases in state-of-the-art climate models. Different models exhibit different biases which lead to rather different mechanisms underlying multidecadal variability in them. An example is given in Fig. 5 for the North Atlantic. Model biases together with the lack of data are reasons that inhibit the exploitation of the full decadal predictability potential in the North Pacific and North Atlantic Sectors. Both, however, offer prospects for improvement.

Fig. 5: Cross correlations as function of time lag (years) between indices of the NAO und the AMOC in control integrations with different climate models (courtesy Jin Ba, IFM-GEOMAR).

References


Alvarez-Garcia, F., M. Latif, and A. Biastoch (2008): On multidecadal and quasi-decadal North Atlantic variability. J. Climate, 21, 3433-3452.

Hurrell, J. W., et al. (2010): Decadal Climate Prediction: Opportunities and Challenges. Sustained Ocean Observations and Information for Society (Vol. 2), Venice, Italy, 21-25 September 2009, Hall, J., Harrison D.E. & Stammer, D., Eds., ESA Publication WPP-306.

Latif, M. et al. (2002): On North Atlantic Interdecadal Variability: A Stochastic View. In: "Ocean Forecasting", N. Pinardi and J. Woods (Eds.), Springer Verlag, 149-178.

Latif, M., and N. S. Keenlyside (2010): A Perspective on Decadal Climate Variability and Predictability. Deep Sea Research, submitted.

Issues In Exploring Mechanisms Of Decadal Variability With CORE-II Forced Global Ocean Models

C. Boening (IFM-GEOMAR) and H. Drange (Uni. Bergen)

Decadal Changes In The Tropical Upper Ocean: Evidence For An Anthropogenic Fingerprint?

L. Terray (CERFACS)

In this talk, I will discuss some of the recent decadal changes observed in the tropical upper oceans. I will focus on the Pacific and Atlantic basins  using two different indicators: surface salinity and upper ocean heat content using an isothermal analysis. I will  then use model results and the detection and attribution framework to identify the causes of the recent decadal changes and in particular, the possible emergence of an anthropogenic fingerprint.

Representing The Meridional Overturning Circulation In Global Eddying Models: Problems and Perspectives

A.M. Treguier (IFREMER)

Sensitivity of simulated Atlantic multi-decadal variability to ocean model formulation and physics

S. Griffies (NOAA GFDL), G. Danabasoglu (NCAR) and S. Yeager (NCAR)

Realistic climate models exhibit a wide range of Atlantic multi-decadal variability characteristics depending on details of the model formulation, grid resolution, and physical parameterization. We present two case studies to illustrate this sensitivity: (A) two climate models where the key difference is the ocean model vertical coordinate (isopycnal versus geopotential); (B)  two climate models where the key difference is the parameterization of overflow processes. These results promote research aimed at understanding how model choices, such as resolution, formulation, and physical parameterizations, impact on the mechanisms affecting simulated Atlantic variability.

A Non-Normal Perspective On MOC Variability And Predictability

L. Zanna (Uni. Oxford), E. Tziperman (Harvard), P. Heimbach (MIT), A. M. Moore (Uni. California)

The variability and the predictability of the Atlantic meridional overturning circulation (MOC) and sea surface temperature (SST) are examined by evaluating optimal perturbations in an ocean GCM with an idealized configuration. Despite the stable linearized dynamics, significant transient amplifications of MOC and SST anomalies occur due to the non-normal dynamics. The mechanisms responsible for the growth of anomalies are analyzed. (i) The MOC is found to be most sensitive to high latitudes deep density perturbations and is reaching a maximum amplification after 7.5 years. This implies that initial conditions and model errors in the deep ocean might limit the predictability of the MOC to less than a decade. (ii) MOC anomalies are found to grow faster when deep ocean perturbations are allowed rather than when only the surface is perturbed such that most predictability experiments in which only the atmospheric state is perturbed may overestimate of the predictability time in the Atlantic. (iii) MOC and SST anomalies growth mechanisms are fully 3D with the participation of westward propagating thermal Rossby waves. (iv) Ocean dynamics is found to participate actively in the amplification of SST and MOC anomalies rather than to passively integrate over the imposed initial forcing such that one may need to consider non-normal effects when inspecting SST trends in the next decades.

Separating A Robust Response To External Forcing From Natural Patterns Of Variability In The Pacific - A. Solomon and M. Newman (CIRES/University of Colorado and NOAA/ESRL)

To what extent decadal climate forecasts will be useful will depend upon whether these forecasts can provide skill on 10-30 year timescales. To assess (a) where this skill comes from and (b) whether the forecast models are simulating realistic variability, it is necessary to be able to separate the natural internal variability from the externally forced variability. This has proven to be a challenge since the forced and internal variability are of the same order on these timescales and the forced variability may be interacting with or projecting on the natural variability (see Solomon et al. 2010). For example, in coupled climate models, the long-term response of the Pacific basin to an increase in greenhouse gases (GHGs) shows some similarities to variability observed during an El Niño/Southern Oscillation (ENSO) event, specifically increased sea surface temperatures (SSTs) in the equatorial Pacific, causing the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment to refer to this variability as “El Niño-like” (Meehl et al. 2007). The similarity between natural internal variability and the projected response to external forcing has made it difficult to determine to what extent changes in natural variability are contributing to long-term climate change in the Pacific.

In this study we present a strategy to separate natural variability that is unpredictable on decadal timescales from a robust response to external forcing that is predictable on decadal timescales. In addition, we show that this strategy is useful in identifying the contribution of natural patterns of variability to the long-term response of the tropical and North Pacific Ocean climate to an increase in GHGs. The results of our study indicate that the total trend is “El-Niño-like”, in that the warming in the central equatorial Pacific is primarily due to variations with three-dimensional structures associated with ENSO. The residual trend pattern has features that can be understood in terms of the “ocean dynamical thermostat” and a cooling in the South Pacific due to increased southeast trades superimposed upon a uniform warming. The residual pattern has greatly reduced within ensemble spread relative to the total trend pattern.

Figure 1: 50-year DJF Pacific ocean temperature (OT) ensemble mean least squares linear trend patterns (left), in units of °C/year, and projection on individual ensemble member fields (right). Trend from A1B 10m temperature fields (A,B), residual 10m temperature fields (C,D), and A1B-residual fields (E,F). Domain means have been removed from the A1B and residual trend patterns to highlight the spatial structure. All values in these two plots are greater than zero. Domain means are A1B=0.021 °C/year and Residual=0.020 °C/year.

Session III: Initialization, Predictability, and Predictions

Limits Of Initial Value Decadal Predictability

G. Branstator (NCAR) and Haiyan Teng (NCAR)

When the climate system undergoes changing external forcing (e.g. from increases in greenhouse gas concentrations), there are two inherent limits on the gain in skill of decadal climate predictions that can be attained from initializing with the observed ocean state. One is the classical initial-value predictability limit that is a consequence of the system being chaotic, and the other corresponds to the forecast range at which information from the initial conditions is overcome by the forced response.

In the first part of this presentation these two limits are quantified for Community Climate System Model, version 3 (CCSM3) with several 40-member climate change scenario experiments. Predictability of the upper-300m ocean temperature, on basin and global scales, is estimated by relative entropy from information theory. Overall, information from the ocean initial condition exceeds that from the forced response for about seven years. After about a decade the classical initial-value predictability limit is reached.

In the second part we quantify the classical initial-value predictability limit for several other models, including GFDL's CM2.1, the Kiel Climate Model, MIROC3.2, and CCSM4. We estimate this limit by examining long control integrations of these models. Significant model-to-model variations in predictability are found.

In the third part we address the question of whether prominent intrinsic modes of the coupled system may have predictability beyond that found for basin-wide measures. In general we find the PDO does not have above average predictability while AMOC predictability lasts a few years longer than basin-wide predictability. However, in CCSM3 and 4, AMOC's predictability is associated with only weak atmospheric signals.

Decadal Variability, Predictability And Prediction: Progress, Challenges And Opportunities

R. Sutton, E. Hawkins, D. Hodson and J. Robson (Uni. Reading)

This talk will provide an overview of recent research related to the subject of the workshop at the UK National Centre for Atmospheric Science (NCAS), and a forward look at challenges and opportunities. There remain fundamental questions about the dominant mechanisms that govern decadal variability in the climate system, and that determine the degree of predictability. There is considerable evidence in support of an important role for the Atlantic Meridional Overturning Circulation (AMOC), but the characteristics of AMOC variability differ considerably between models, and there is a need to understand the reasons at a mechanistic level. One dimension that we have been investigating is the sensitivity of AMOC variability to climate model resolution. We are also researching predictability. The extent to which internal climate variability is predictable beyond seasonal (ENSO) timescales is still a matter of debate. Whilst some large scale ocean variables may be predictable several years ahead, there is little evidence that surface climate variables are significantly predictable more than a year or, at most, two ahead. The development of decadal predictions presents yet more challenges, notably in relation to initialisation and the design of ensembles to sample uncertainties efficiently. We have been investigating the potential of singular vectors as a tool for identifying which ocean observations may be most valuable for constraining decadal predictions, and for informing ensemble design. Lastly, in collaboration with the Met Office Hadley Centre, we are exploring issues related to the use of anomaly initialisation for decadal predictions, and how best to exploit the new AMOC observations from the 26N RAPID array.

Decadal Prediction Strategies: Recommendations And Illustrations From CLIVAR EASYINIT

W. Hazeleger (KNMI)

In this presentation the main recommendations of the CLIVAR workshop on Earth System Initialization for Decadal Predictions will be presented (see http://www.knmi.nl/samenw/easyinit/ ). At this workshop, strategies for initialisation, perturbation and verification were discussed. The recommendations will be presented and illustrated with results from EU Ensembles near term predictions and decadal predictions made with EC-Earth using NEMOVAR initial ocean conditions.

For initialisation full and anomaly initalisation strategies are proposed. A comparison of the strategies will be shown with emphasis on the drift that is generated. It will be shown that even when using anomaly initialisation drift corrections are needed and the linearity of drift will be discussed. For perturbation of the initial conditions to generate the ensemble, a range of options have been discussed. Especially perturbation of the deep ocean remains an open issue. For verification some clear recommendations were made. Best practices, such as comparison against ‘own’ analyses, test against simple statistical models and drift corrections were recommended. Examples will be given by verifying the multi-model near term predictions of the EU Ensembles project. These predictions show skill on decadal time scales in the North Atlantic.

Finally, some practical recommendations were made, such as the collection of ocean analysis sets. The University of Hamburg is currently setting up a central database for the community. Also, coordinated model experiments were recommended.

Recent Progress In Ocean Reanalysis And Initialization At ECWMF

M. Balmaseda, Linus Magnusson, Kristian Mogensen and Franco Molteni (ECMWF)

A new ocean reanalysis with NEMOVAR has been produced at ECMWF. It will be used to initialize the EC-EARTH decadal forecasts. The climate variability in the NEMOVAR reanalysis is evaluated by conducting sensitivity experiments. In particular, the sensitivity of the Atlantic Meridional Circulation and trends in steric height to the choice of assimilation parameters and multivariate formulation is discussed.

The NEMOVAR ocean reanalysis has been used to test different initialization and forecast strategies for seasonal and decadal predictions with a recent version of the ECMWF coupled model. Results from coupled forecasts using full initialization, anomaly initialization and flux correction are presented.

This work contributes the EU-funded project COMBINE. The COMBINE activities on initialization of the decadal forecasts will be outlined.

A Metrics Framework To Assess And Validate Decadal Climate Predictions And Simulations

L. Goddard (IRI), on behalf of the US CLIVAR Decadal Predictability Working Group

The US CLIVAR Decadal Predictability Working Group has as one of its main objectives to define a framework for forecast verification. Given the early stages of the framework development, this presentation will outline what such a framework might include. The experience gained from the production, communication, and application of more than two decades of seasonal-to-interannual prediction serves as a highly relevant starting point, but it is recognized that decadal prediction contains some unique elements. The working group intends to use the opportunity of this presentation to solicit feedback from those at the workshop to strengthen and solidify the guidance offered by the verification framework. The intended beneficiaries of this guidance would span the range from modeling centers to end-users. It should, therefore, be a careful balance between highlighting what successes are achieved by the predictions, and managing expectations of their potential use. The framework should also better prepare modeling and prediction centers to understand and communicate the quality of the information they have to offer, as they start to provide decadal predictions to the scientific community and the public at large. Similarly, users of the predictions require information on forecast quality, relative to baseline information they already utilize in their decision making context, e.g., the climatology, as well as information about the limitations and caveats of decadal forecasts.

Decadal Prediction Of North Atlantic Hurricane Frequency

D. Smith, R. Eade, N. Dunstone, D. Fereday, J. Murphy, H. Pohlmann and A. Scaife (UK Met Office)

North Atlantic hurricane activity has increased substantially since the 1970s, but whether this was natural internal variability or externally-forced, remains unknown. Either way, hurricane frequency is potentially predictable because climate models can directly simulate year-to-year variations in Atlantic tropical storm frequency if forced by observed sea surface temperatures, and decadal climate predictions are now being made that are initialised with the observed state of the climate system to predict both internal and externally-forced changes. However, skilful predictions have previously been limited to a season ahead, and previous evidence for external forcing of hurricanes is indirect, relying on statistical relationships or establishing external influences on environmental factors that are believed to affect hurricane development. We present results from a comprehensive set of decadal hindcasts covering the period from 1960 to 2005, showing skilful predictions of Atlantic tropical storm frequency well beyond the seasonal timescale. Results from a parallel set of uninitialised hindcasts suggests that the recent increase was not caused by internal variability alone. Initialisation of our model with the observed state of the climate improves forecast skill, mainly through improved predictions of ocean conditions in the tropical Pacific and north Atlantic that influence hurricanes via established physical mechanisms. This provides further evidence of remote ocean forcing, and highlights regions where continued monitoring is important.

Predicting And Projecting Hurricane Activity Changes

G. Vecchi (NOAA GFDL)

We explore recent progress towards predicting and projecting tropical storm and hurricane activity changes using dynamical and statistical models. We outline a strategies for regional climate change prediction as applied to hurricanes, and discuss aspects of the hurricane prediction problem that make it particularly challenging. In order to predict changes in hurricane activity, one must: i) define a measure of activity, ii) test prediction methods on historical cases (making efforts to correct historical estimates of this change in activity for data problems), iii) develop comprehensive dynamical models and theoretical understanding to connect changes in activity to large-scale environmental parameters.

Dynamical models and simple statistical models exhibit skill in interannual predictions of Atlantic and East Pacific hurricane counts. Dynamical predictions and projections of hurricane activity changes are found to be critically sensitive to details in the horizontal structure of SST changes and the vertical structure of atmospheric temperature changes - rather than the overall temperature change. The pattern of SST change provides a unifying statistical framework for seemingly divergent hurricane projections using dynamical models. The sensitivity of hurricane activity to details in the response of the climate system limit their predictability. Efforts are made to identify the most influential patterns of SST change, and connect these relevant patterns to forced and internal climate variations.

Recent increase in Atlantic TS and hurricanes can be forced by some estimates of historical SST changes, yet hindcasts of hurricane counts using comprehensive dynamical models are limited by the observational uncertainty in SST - even over the satellite era (1982-2007). Thus, in order to make accurate decadal predictions of hurricane activity changes, we may need to predict decadal patterns of SST change better than we currently observe them.

On North American Decadal Climate For 2011-2020 - M. Hoerling (NOAA ESRL), A. Kumar (NOAA CPC), J. Hurrell (NCAR), Xiaowei Quan (NOAA ESRL)

There are many important challenges and questions associated with the notion of climate prediction on decadal time scales. For instance, given imperfect and incomplete observations and imperfect models, what is the best method of initialization and what effect does initialization have on the skill of climate predictions? Ultimately, the predictability is a function of the mechanisms responsible for the variability, and many fundamental questions remain regarding the nature of and processes responsible for decadal variations in climate. This study examines the evidence for decadal variability in North American climate, assesses plausible mechanisms, and explores prediction and predictability of the continent's climate for the upcoming decade.

We use dynamical modeling methods to develop plausible scenarios of North American climate for 2011-2020. The underlying principle guiding the investigation is that North American decadal climate conditions are sensitive to the state of global surface boundary conditions, particularly sea surface temperatures. Our method involves generating plausible global SST and sea ice scenarios for the coming decade, conditions that are then specified in suites of un-initialized AGCM integrations in order to determine the climate responses and their uncertainties.

To motivate the design of model experiments for 2011-2020, we first address the extent to which annual North American temperature and precipitation since 1900 exhibits decadal variability, and address their causes and potential predictability by comparing with CMIP3 and AMIP simulations of the last century. These reveal the important role played by the trend components in global SSTs and sea ice, and the analysis also indicates that appreciable decadal variations in North American climate can result from decadal SST variability.

Three aspects of the multi-model, large ensemble AGCM experiments for 2011-2020 are highlighted. One is the sensitivity of responses to various, plausible scenarios for the change (trend) component in global SSTs. Second is the impact of internal decadal SSTs with a focus on extreme phases of the so-called Pacific Decadal Oscillation (PDO) and Atlantic Multi-decadal Oscillation (AMO). Third, we address the amplitude of North American decadal conditions resulting from unforced atmospheric circulation variability and assess potential predictability via classic signal-to-noise ratio methods.

Decadal Predictions with EC-Earth - First Results

B. Wouters, G. J. van Oldenborgh, W. Hazeleger (KNMI)

A joint EU project between 20 institutions in 9 European countries, THOR stands for "Thermohaline Overturning – at Risk?",and aims to establish an operational system that will monitor and forecast the development of the North Atlantic thermohaline circulation on decadal time scales and assess its stability and the risk of a breakdown in a changing climate. Through the assimilation of systematic oceanic observations at key locations into ocean circulation models it will provide a set of geo-observational products that will be used to forecast the development of the system using global coupled ocean-atmosphere models.

We focus on the decadal predictability of the overturning circulation and associated climate variables, for which the EC-Earth v2.2 model is used. The model configuration used consists of the ECMWF IFSc31 model at T159/L62 resolution, the NEMO2 ocean model at 1 degree resolution and the LIM2 sea-ice model. Analysis of a ~500 year run with pre-industrial conditions shows significant potential predictability in, e.g., sea surface temperature, especially in the North-Atlantic, even after correcting for linear trends in the data.

To assess the actual skill in predictability of the model, a hindcast experiment is carried out. Each 5 years, starting 1 Nov. 1960, a ten year model run  consisting of 5 members is launched according to the CMIP5 protocol. A full initialization method is used, where the ocean is initialized using NEMOVAR re-analysis data and sea-ice is obtained from a forced ocean-only model run. In this presentation, we present the first results of this decadal prediction experiment. As expected, the model shows a drift in the first years due to the initialization shock from the full initial state, but stabilizes afterwards. After correcting for this drift, reasonable to good agreement is found between the hindcasts and the ocean re-analysis data in, e.g., North Atlantic sea surface temperature and overturning strength.

First Results Of CMIP5 Hindcasts Experiments At IPSL - J. Mignot, S. Labetoulle, D. Swingedouw, E. Guilyardi, S. Masson and G. Madec (LOCEAN, IPSL)

First results of the CMIP5 "Near term" IPSL-CM5 (NEMO ocean / LMDz atmosphere) decadal predictability experiments are presented. The strategy chosen at IPSL is to initialize the climate model only at the ocean-atmosphere interface, following the guidance and expertise gained from ocean-only NEMO experiments. Different solutions of surface nudging will be presented, evaluated and compared to previous studies. In particular, we show that the model is rather successfully initialized globally by SST nudging only when the nudging strength is correctly tuned. The effect of imposing the observed wind stress on top of SST nudging will also be discussed. A second, novel and more physical, approach for initializing the climate system consists in computing the heat, salt and momentum fluxes received by the ocean model as a linear combination of the fluxes computed by the atmospheric model and by a CORE-style bulk formulation using up-to-date reanalysis. Results from the two approaches will be compared.

Session IV: Ocean and Coupled Synthesis

Ocean Syntheses: Status And Limitations

P. Heimbach (MIT) and R. Ponte (AER Inc.)

Synthesis efforts that provide a description of the global ocean state and its temporal evolution are essential for climate analysis and prediction. Here we attempt a brief review of the subject, drawing considerably from experience with our own synthesis efforts as part of the ECCO* project. Various state estimates of near-global extent and covering periods dating back several decades are available, but prior to the altimeter and Argo era, data is very sparse in space and time. Until the early 1990's, available estimates are thus very weakly constrained. Data thinning experiments suggest that observations in the modern period (Argo, altimetry) provide essential constraints for determining the ocean state. Even for the modern period, however, severe under-sampling remains, particularly at depth (< 2000 m), at high latitudes including most ice-covered areas, and most of the marginal and shallow seas. Most data sets are not dense enough to provide good constraints on the scales relevant for climate analysis, particularly in regions with large eddy energy. On decadal timescales, all these shortcomings become important and a truly global observational database, reaching all the way to the bottom, is warranted. As for the model used for the synthesis, resolutions continue to improve, but there is still need for the implementation of better parameterizations of climate-relevant processes (interior mixing, deep convection at high latitudes, density-driven flows across sills and off shelves, etc.). Synthesis efforts that combine all data types (in situ: moorings, drifters, Argo, etc.; satellite: altimetry, SST, gravity, salinity, winds, etc.), allow parameter tuning and bias corrections (e.g., internal mixing coefficients) to improve the treatment of model errors, and provide estimates consistent with atmospheric boundary conditions and conservation principles of momentum, heat and freshwater, offer considerable advantages. The lack of a full description of uncertainties, both a priori and a posteriori, in the available ocean state estimates is a severe limitation for their use in climate analysis and prediction. Methodologies for provision of realistic error estimates need to be implemented. The hightened interest in high-latitude processes requires adequate treatment of sea ice, along with the use of sea ice and, where available, under-ice observations. On the decadal time scales, developing coupled atmosphere-ocean estimation systems is of considerable relevance to overcome serious limitations with existing atmospheric re-analysis products, all of which suffer from large heat and freshwater imbalances, and which compromise accurate budget calculations.

*Estimating the Circulation and Climate of the Ocean

Errors In Ocean Syntheses: Estimation And Impact - A. Koehl (Uni. Hamburg)

One of the key activities of the CLIVAR Global Synthesis and Observations Panel (GSOP) is the intercomparison of current ocean syntheses. Although intercomparison provides a first step in assessing quality and usefulness of synthesis products for climate analysis and prediction, it remains an insufficient instrument and ocean measurements are in general required for evaluation. Error estimates for ocean syntheses are needed if it comes to key quantities that are not accessible to direct measurement such as transports but also in general for the initialization of climate models. Because the model systems used for analysis and for prediction are often different, initialization requires a blending step that transfers the information from the analysis into the prediction system. Although for this step the error of both systems needs to be known, this is often neglected and syntheses are treated as perfect. We provide ideas about the error associated with ocean syntheses  and demonstrate its impact on the error of transport estimates including ideas where the synthesis errors need to be reduced. This is a step towards understanding differences among the various syntheses.

Decadal Hindecasts, Predictions And Predictability At GFDL

T. Rosati (NOAA GFDL)

Two Tales Of Initializing Decadal Climate Predictions With The ECHAM5/MPI-OM Model - D. Matei (MPI-M), H. Pohlmann (MPI-M, UK Met Office), J. Jungclaus (MPI-M), W. Mu.ller (MPI-M), H. Haak (MPI-M) and J. Marotzke (MPI-M)

We investigate the forecast skill of decadal climate predictions using two different ocean initialization strategies. First we use the ocean synthesis data provided by the GECCO project (Köhl and Stammer, 2008) as initial conditions for the coupled model ECHAM5/MPI-OM. Ten year- long hindcast experiments are then performed over the period 1952-2001. An alternative approach is one in which the ocean temperature and salinity are initialized from an ensemble of ocean model runs forced by the NCEP-NCAR atmospheric reanalyses for the period 1948-2007. An anomaly coupling scheme is used in both approaches to avoid hindcast drift and the associated initial shock. Differences between the two assimilation approaches are discussed by comparing them with the observational data in key regions and processes. We asses the skill of the initialized decadal hindcast experiments against the prediction skill of the noninitialized hindcast simulations and statistical forecasts. The predictive skill of the GECCO and NCEP-forced initializations is regionally similar; we obtain significantly improved prediction skill for sea surface temperature up to a decade ahead over the North Atlantic, the Western Pacific/Indian Ocean, and the Mediterranean. Skilful predictions of surface air temperature are found for the second pentade (yr6-10) over

Western Europe, Northern Africa and Central and Eastern Asia. The hindcasts initialized from the NCEP-forced run give a higher skill during the first pentade, while the hindcasts started from GECCO have a higher skill in the second pentade. We also analyze the potential predictability of the Atlantic Meridional Overturning Circulation (AMOC). Since direct measurements of MOC are limited to a few years, we compare the predicted values to the respective assimilation experiments. Hindcasts of Atlantic MOC show higher predictability than the comparison experiments without initialization and damped persistence predictions up to about 6 years. The NCEP hindcasts show potential predictability of the two main components of the North Atlantic Deep Water (Nordic Seas Overflows and Labrador Sea Water) up to 5 years in advance.

Sea-Ice Initialization In EC-EARTH - K. Wyser (SMHI), M. Caian (SMHI), T. Königk (SMHI), C. Jones (SMHI), and C. König Beatty (Université Catholique de Louvain)

How can sea-ice be initialized for decadal predictions? Sea-ice cover is available from satellite observations, at least for a few decades, but observations of sea-ice thickness are scarce. Furthermore, the state of the sea-ice is closely related to the ocean state and cannot be specified independently. The initialization of sea-ice thus requires some special attention and we have tested several methods how sea-ice can be initialized in climate models for decadal climate predictions.

All our simulations were done with the EC-EARTH model, a newly developed coupled climate model that is used for CMIP5 experiments. It comprises NEMO as the ocean model with LIM2 as its sea-ice component and an option to replace LIM2 with LIM3.

For decadal simulations with EC-EARTH, the ocean is initialized from a NEMOVAR re-analysis that is provided by ECMWF. For the sea-ice we have tested the initialization of sea-ice fraction and thickness from climatology. The method yields reasonable results in decadal simulations, but attention needs to be paid to the coherence between ocean and sea-ice state

Another method for the initialization of sea-ice is to use the results from a forced run of the ocean model. The sea-ice from such a run can be used to initialize a decadal simulation of the coupled model, either directly in a full-field initialization or in a setup with anomaly initialization. Both methods are tested in a number of decadal simulations of the recent past.

Finally, we plan to use the sea-ice from a forced NEMO run that uses an Ensemble Kalman filter to data assimilate sea-ice observations. This method will provide sea-ice cover and thickness that are close to observations yet still retain a coherent structure with the ocean model.

Data requriements for ocean reanalyses

D. Stammer (Uni. Hamburg)